Mattermost Workflows for Codex
A local-first Codex plugin that bundles an MCP server to safely interact with Mattermost, enabling channel/thread summarization, conversation search, and reviewed post publishing via the Mattermost REST API.
README
Mattermost Workflows for Codex
A local-first Codex plugin for working safely with Mattermost through a bundled MCP server. It uses the standard Mattermost REST API v4 and is compatible with Mattermost 10.11, so it does not require the official Agents MCP plugin or administrator access.
Codex can summarize channels and threads, search conversations, and publish a reviewed post. The MCP process runs locally and connects directly from the user's machine to the user's Mattermost deployment. The publisher never receives that traffic.
This project is not affiliated with, endorsed by, or sponsored by Mattermost, Inc. or OpenAI.
MVP
- Identify the connected user and list accessible teams.
- Resolve team and channel names using the account's existing permissions.
- Read and summarize a channel or thread.
- Search posts within a requested team or channel scope.
- Freeze a new post or reply, show the exact target and text, and publish it only after explicit approval.
- Store local credentials outside the repository, with macOS Keychain support.
Creating users, teams, or channels; membership changes; impersonation; attachments; bulk posting; background indexing; and browser-cookie reuse are outside the MVP. See MVP scope.
Requirements
- Mattermost 10.11 or another deployment exposing the compatible REST API v4 endpoints.
- A normal Mattermost account with access to the channels you want to use. No admin access is required.
- Codex with plugin and local MCP support.
- Node.js 18 or later only when building from source. The plugin includes the built server.
Install
From this repository marketplace:
codex plugin marketplace add shanginn/codex-mattermost --ref main
codex plugin add mattermost-workflows@personal
Start a new task after installation so Codex loads the plugin and its MCP server.
Connect your Mattermost account
On macOS, the recommended setup creates an ordinary Mattermost session and stores only the resulting token in Keychain. Run this from the cloned repository and substitute your own host and login locally:
node plugins/mattermost-workflows/scripts/configure-macos.mjs \
--server-url https://chat.example.com \
--login-id YOUR_LOGIN
The password is entered interactively, is not printed, and is not saved. The generated config lives at ~/.config/codex-mattermost/config.json; it contains the server URL and Keychain lookup metadata, but no secret.
You can alternatively supply an existing session token or personal access token through standard input, or use the MATTERMOST_URL and MATTERMOST_TOKEN environment variables. Never commit either value or paste secrets into chat. See the complete setup guide.
Example prompts
- “Who am I connected as, and which Mattermost teams can I access?”
- “Summarize decisions and action items from the channel I name since Monday.”
- “Find posts in this team about database connection exhaustion.”
- “Draft a maintenance notice for the channel I name. Do not post it until I approve.”
Development and tests
The plugin lives at plugins/mattermost-workflows; the local marketplace is .agents/plugins/marketplace.json.
npm install
npm test
The test suite builds the bundled server, exercises the MCP protocol and Mattermost API adapter against local fixtures, and validates the plugin, bilingual documentation, safety gates, attribution, and reviewer cases. It never connects to a real Mattermost deployment.
Privacy, support, and terms
- Privacy · Конфиденциальность
- Terms · Условия
- Support · Поддержка
- Contributing · Участие
- Third-party notices
License and required attribution
This project uses the custom Shangin Attribution License 1.0. Any use, deployment, derivative, or fork must include:
Based on Mattermost Workflows for Codex by Nikolai Shangin (shanginn@gmail.com).
The custom license is source-available and attribution-required; it is not represented as an OSI-approved license. Third-party components remain under their own licenses.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。